R E V I E W

A R T I C L E S

A N D

COMMENTARIES

Family Mediation Research: Is There Empirical Support for the Field?

JOAN B. ICELLY

hen family mediation emerged on a national level twenty-five years ago, divorce mediation was promoted as less expensive and timeconsuming, more humane and satisfying to participants than litigation, resulting in better compliance with agreements and reduced relitigation. It was expected to enhance problem-solving skills among the disputants, promote cooperation and communication regarding children, and result in better adjustment of adults and their children to the divorce and be more empowering of the participants than traditional adversarial processes. Research was initiated in public and private sectors to assess these claims and to respond to the strident challenges and opposition to divorce mediation that quickly emerged. Lawyers were openly skeptical that the process was more fair and less costly. Women's advocates, who questioned women's ability to negotiate equitable arrangements with men, made policy pronouncements advising women against participating in mediation. Ironically, these claims were made in the absence of empirical studies assessing the fairness and impact of adversarial divorce processes on participants, including those with a history of domestic violence. Early research focused on the basic questions of settlement rates, client satisfaction, time and cost efficiencies, comparison of outcomes, compliance, and, to a lesser extent, impacts on parental confiict, communication, cooperation, and psychological adjustment (see Beck and Sales, 2000; Emery, 1994; Kelly, 1996). This first-generation research, conducted in the United States, Canada, England, and Australia, constitutes the largest body of empirical research among any of the mediation sectors. Variations in research populations, methodologies, measures, and dispute settings have been the norm, making it problematic to generalize about family mediation or rely on a single study. Many research publications failed to provide
CONFLICT RESOLUTION QUARTERLY, vol. 22, no. 1-2, Fall-Winter 2004 Wiley Periodicals, Inc., and the Association for Conflict Resolution 3

ICELLY

basic descriptors, such as the nature of the population served, number of sessions and hours of service, the model (if any) mediators used, and whether premediation screening was used. Legal rules and cultural contexts ofthe jurisdictions that might affect outcomes were rarely described. Despite these problems, convergence on many questions has emerged over two decades, indicating that some major findings regarding family mediation are robust and replicable across settings. This article summarizes a selected group of family mediation studies published over the past twenty years, with a focus on four custody mediation programs in the public sector, two studies of public and private sector comprehensive divorce mediation, and three court-connected programs for mediation of child protection or dependency disputes. Studies selected for inclusion had adequate sample sizes, objective measures, appropriate statistical analyses, and, when available, comparison groups and longitudinal designs. The primary structural elements described in the family mediation research have variously included the timing of intervention, whether it is voluntary, who pays, the level of self-determination, and the training of neutrals. The criteria used to determine the success most often have been settlement rates, satisfaction of participants, efficiencies in time and cost, and, to a lesser extent, evidence of changes in relationships and durability of settlement. Because most mediation in the custody and divorce family sector occurs between male and female parties without lawyers present, the issue of who can mediate has been an important one, and the nature of the disputes, domestic violence, and characteristics of the parties is described as available. It should be noted that all group or gender differences were statistically significant.

Custody Mediation in the Public Sector

Research on child custody mediation in the public sector has been concentrated primarily in California, Virginia, Canada, and several multiple jurisdictional studies.
Custody Mediation Research in the California Courts

California legislation mandated mediation for all separating parents with custody and visiting disputes, effective 1981. The Statewide Family Court Services, now the Center for Families, Children and the Courts (CFCC), initiated in 1991 a series of interlocking studies that provide a unique uniform statistical database consisting of representative and longitudinal

Family Mediation Research

data from nine data collections involving over eighteen thousand contested child custody cases that used mediation. Because mediation was mandatory, there was no litigation comparison group. These snapshot studies have described the population, services provided, case complexity, the usefulness of mediation for contested child disputes, and the participants' feedback about their experiences.' Structural Elements ofthe Statewide Program. Custody mediation is an early intervention for disputing parents, who are required to schedule a mediation appointment within several weeks of filing a motion or petition. Attendance at one mediation session is mandated, and services must be provided free of charge. Subsequent legislation provided for separate sessions, opt-outs for parents, and special assessment for families where domestic violence was alleged or had occurred. When domestic violence was raised as an issue, in 57 percent of cases, parents attended separate mediation sessions (Center for Families, Children and Courts, 2002). Parents can return to custody mediation following judicial settlement conferences and following divorce or final orders. Parents can opt to use private fee-for-service custody mediation, or settle through private ordering, but the majority of disputing parents use family court services. In 1996, 84,550 mediation cases were seen, and 26 percent of these were seeking a modification of existing orders (Center for Families, Children and Courts, Report 12, 2000). Currently, in thirty-four of California's fifty-eight counties (C. Depner, personal communication, Dec. 8, 2003), mediators are authorized to make recommendations to the court for custody and visitation when parents are at impasse, whereas in the remaining counties, mediation is confidential. Thus, in "recommending" counties, mediation incorporates an evaluative component that most likely shapes the style of mediation offered. The extent of self-determination varies according to whether parents are able to reach partial or full agreement in mediation, must rely on either mediator recommendation or other more formal evaluative processes to have the issues resolved, or when allegations of child abuse or violence require temporary orders. California statute provides that court mediators have a masters degree in family counseling or behavioral sciences. The number of mediators with extensive or formal mediation training is unclear, although all staff attend an annual three-day conference for mediators and family court judges. Mediators are salaried by the state, with the exception of those in rural counties with small populations, who serve on a contractual basis.

KELLY

Participants in mediation have a right to counsel, but attorneys are rarely included in the mediation itself In 53 percent of court family mediation cases, at least one parent is not represented, a trend that has increased over the years and is related in part to the lower socioeconomic status and higher unemployment rates of mediation participants, compared to statewide means (Center for Families, Children and Courts, 2000). The court custody mediation population is quite diverse: 48 percent white, 29 percent Hispanic, 8 percent African American, 4 percent Asian/Pacific Islander, 2 percent Native American, and 5 percent- other. In contrast, 80 percent of custody mediators identify themselves as non-Hispanic white (Center for Families, Children and Courts, 2001). Despite efforts to hire bilingual mediators and recruit interpreters (one county had forty-seven languages represented among parents), language and cultural barriers remain a ' challenge. The large numbers of parents reporting domestic violence present additional challenges. In the CFCC snapshot study, interparental violence was reported by at least one parent in 76 percent of the twenty-five hundred mothers and fathers. Sixteen percent of mothers and 8 percent of fathers reported violence during the prior six months. In 47 percent of mediated cases, at least one parent reported on the (independent) survey that violence had ever occurred, but neither parent raised domestic violence as an issue before or during the session. Outcomes of Custody Mediations in California Courts. In California, custody mediation research reports on outcomes such as settlement rates, satisfaction, and changes in participant relationships. Settlement rates. In a two-week period of 1991, the snapshot study of 1,388 families in 1,388 mediation sessions (comprising 82 percent of all mediations during that period) reported that 55 percent of the families reached agreement. Among the 45 percent who did not reach agreement, more than one-quarter were scheduled for fiirther mediation, but the number of those who settled in later mediations was unknown. The context for this settlement rate is that 48 percent characterized their interparental conflict as high, and more than half raised at least one serious issue (for example, child neglect or abuse or substance abuse). The most common issues raised were domestic violence (in 29 percent of cases), substance abuse (25 percent), and maligning (21 percent) or harassing (19 percent) the other parent. More than half of disputants were or had

Family Mediation Research

previously been protected by domestic violence restraining orders, and one-third reported serious concerns about a child's emotional well-being. On a level-of-diflSculty scale (from 1 to 10), 2,812 mediators in the 1999 CFCC snapshot study rated 23 percent of their cases as a 9 or higher and 39 percent as an 8 or higher. Cases were rated higher in difficulty when one (32 percent) or both (26 percent) parents raised multiple serious issues, and at least half of difficult cases (rated 8-10) did not reach agreement in one mediation session (Center for Families, Children and Courts, 2003a). CFCC has recently decided that a single settlement rate statistic is no longer useful because of variations in case complexity, different services available in various courts, and legislative mandates requiring special assessments in domestic violence cases. Satisfaction of participants. The majority of over sixty-seven hundred mothers and fathers in the 1991 and 1993 snapshot studies reported substantial satisfaction on sixteen aspects of the mediation process and the outcomes (Statewide Office of Family Court Services, 1992; Center for Families, Children and Courts, 1993a). More than 90 percent of clients said the mediator explained mediation procedures clearly, treated them with respect, listened to their concerns, and tried to keep them focused on their children's interests. Seven in ten clients said the mediator helped them to see more ways to work together as parents. However, 13 percent of mothers and fathers complained that they felt pressured by the mediator to go along with things they did not want, and up to 17 percent felt rushed by the mediator. The percentage of dissatisfied clients is consistent with other studies. Overall, 86 percent of parents said they would recommend mediation to a friend. Cender differences appeared on four of sixteen indicators of client satisfaction. More women than men felt that the mediator listened to their concerns, and fathers were more likely to feel that mothers had an unfair advantage in mediation. Women were more likely than men to report that they felt too intimidated to say what they really felt in the 1991 study, but this difference was not replicated in the 1993 study (Center for Families, Children and Courts, 1993a). Subsequent analyses suggested that feelings of intimidation were related less to gender dynamics than to clients' education and financial resources. Ethnic and income differences in satisfaction also emerged, with mediation seen as more helpful in several dimensions by parents with less education and lower income and by ethnic minorities (Center for Families, Children and Courts, Report 3, 1994). The issue of mediator recommendation to the court continues to be a controversial one among outside observers, but not apparently for the

KELLY

parties. Seventy-four percent of parents in "recommending" counties versus 78 percent of parents in "nonrecommending" (confidential) counties were satisfied with the process and outcomes, a nonsignificant difference. Reaching an agreement in mediation was the stronger determinant of client satisfaction, with client ratings overall approximately 12 percent higher when compared to those who remained at impasse and therefore had to use other more adversarial court processes to get their disputed issues resolved (Center for Families, Children and Courts, 1994b). Changes in participants' relationships. Follow-up studies with 1,532 parents two years later indicated a decline in measures of satisfaction, including whether mediation was a good way to come up with custody and visiting plans (93 versus 67 percent) and whether the agreement was perceived to be a fair one (87 versus 68 percent). Satisfaction remained higher among those who reached agreement, compared to clients who did not reach agreement in mediation and used other court processes, particularly for mothers. Fathers using other court processes were the least likely to view the results as fair (55 versus 63 percent of fathers reaching mediated agreements). On ratings of severity of family problems, such as violence and abuse, there were no differences between the two-parent groups, which might account for these different levels in satisfaction (Center for Families, Children and Courts, 1994b). A more recent study (Center for Families, Children and Courts, 2000) of client feedback indicates that satisfaction rates remain high across many measures of process, with participants with lower education and income, and being in pro per, more satisfied than better-educated, higher-income parents. Follow-up studies of 1,532 parents two years after the 1991 data collection found that 51 percent of parents who did not reach mediated agreements reported that the way they arrived at custody agreements (mediator recommendation, judicial settlement, or custody evaluation) had a negative effect on their relationship with the other parent, compared to 42 percent of those mediating their agreements, a significant difference. And 55 percent of parents using mediation to reach agreements said the procedure helped them to work together as parents, compared to 34 percent of those using other legal processes. Outcomes. The range of agreements resulting from mediation is comparable to those in the general divorcing population, with the most common outcome being joint legal custody and mother physical custody (27 percent of families). Although critics of mediation have claimed that women are forced to agree to joint physical time-sharing arrangements, the California court data do not support that view, nor do the data already cited regarding

Family Mediation Research

women's satisfaction with the mediation process (see also Kelly and Duryee, 1992). Sixteen percent of parents chose a joint time-sharing arrangement (defined as eight to twenty overnights with the mother in a four-week period), which was most often linked to experience with de facto custody arrangements prior to mediation (Center for Families, Children and Courts, 1994a). Compliance with agreements. Mediation parents more often reported that the mediated agreement had sufficient detail to guide them, compared to those using other court methods {GA versus 53 percent), and 55 percent of parents did not report difficulties in compliance with their agreements. Among the 45 percent with problems reported in sticking with the visitation schedule, there were no differences between parents who mediated agreements and those using other court processes.
Mediation of Parenting Time and Responsibilities in a Colorado Court

In order to determine if mediation of parenting disputes produced measurable benefits to the Colorado courts, ninety-two mediation cases were compared to one hundred cases that had not participated in mediation (Thoennes, 2002a). Structural Elements ofthe Colorado Project. Research was conducted in the Tenth Judicial District, where mediation of domestic relations disputes was the largest component of alternative dispute resolution (ADR) services offered. The mediation cases were systematically referred by the court in the early stages ofthe dissolution (generally within forty-five days of filing) during 1999-2000; the comparison cases in the study had filed for dissolution during 1996-1997 and had child-related disputes but either did not attend mediation (the majority) or did so only at a much later stage (typically five or more months after filing). Waivers to exempt parents from mediation could be obtained for domestic violence or where participation would be a hardship. The final research sample had ninety-two families in the mediation group and one hundred in the comparison group. Baseline comparisons of the two samples indicated no significant differences in income, length of marriage to filing, and number or age of children. Indicators of disputes about children were similar in both groups. The training and qualifications of the mediators are not reported, nor was type of mediation offered and role ofthe mediator described. The program charged fifiy dollars per party per hour, unless parties were indigent (the fee was waived) or low income (the fee was halved). Parties had a right to counsel. Thirty-four percent of the mediation and 24 percent of the comparison cases had at

10

KELLY

least one parent without legal representation, more often father. Most cases were handled in one session of approximately two hours. Outcomes of Mediations in Colorado Tenth Judicial District. The outcomes studied in the Colorado research focused less on participants' perceptions and more on effectiveness of dispute processing. Settlement rates. In 39 percent of cases, the parents reached full agreement on all issues, an additional 55 percent reached agreement on some issues, and 6 percent reached no agreements. Families reaching partial agreement returned to the court with stipulations regarding the settled disputes and had hearings to resolve the rest of the disputed issues. Satisfaction and changes in relationships. These were not assessed in this study. Efficiencies in dispute processing. On the first outcome criterion, the amount of time that cases were open, the mediation group had significantly fewer days between filing and final orders (the mean was 334) than did the comparison group (the mean was 395). On the second, stipulated agreements, the mediation group was significantly more likely to present stipulations on parental decision making, child residence, and child support than did the comparison group, which saved judicial and court time. On additional criteria, the mediation group was significantly less likely than the comparison group to have a continuance (15 versus 31 percent), filed fewer motions than the comparison group cases (means of 2.4 versus 3.6 motions filed), and more mediated cases filed no motions (21 versus 8 percent). Twelve percent of mediated casesfiledfiveor more motions, whereas 25 percent ofthe comparison groupfiledfiveor more motions. Finally, the amount of hearing time scheduled was significantly less for the mediation group than the comparison group (2.0 hours versus 3.6 hours), and only 5 percent of mediation cases had hearings scheduled for a full day, compared with nearly 25 percent of the comparison group. Thus, mediation produced a more efficient and timely fiow of cases, with fewer unexpected delays and motions. Relitigation. Rates of relitigation are often cited as a measure of mediation effectiveness.^
CharlottesvUle Custody Mediation Project

Parents filing a petition for a custody or visitation hearing in a central Virginia court between 1983 and 1986 were randomly assigned to mediation or litigation and invited to participate in a study. Few differences were

Family Mediation Research

11

found between those who agreed or refused to participate or between the two groups, who were predominantly young, working-class parents. Excluded from the mediation group were residents at a battered women's shelter or when a domestic violence or child abuse charge was pending. Some participants indicated during the mediation that they experienced violence earlier in the relationship (R. Emery, personal communication, Dec. 9, 2003). Thirty-six families proceeded through the usual court procedures and thirty-five families entered the mediation service. In a replication study, fifteen mediation and sixteen litigation families were combined with the original sample, resulting in a sample size for follow-up studies, after attrition, of seventy-one families. This is the only random assignment study of the relative effectiveness of custody rnediation, and despite the small sample size, it merits attention (see Emery, 1994). Structural Elements in the CharlottesvUle Study. Participants with custody or visitation disputes were recruited immediately prior to the initial court hearing. For the mediation families, mediation was an early voluntary intervention, whereas the litigation group proceeded through the usual court processes, typically using attorneys. Mediations were conducted by pairs of male and female comediators, all of whom had at least a master's degree in a mental health field and training in mediation. Mediation was free of charge and limited to six two-hour sessions. Participants completed mediation in an average of 2.4 meetings. All participants in the study had a right to counsel. Descriptions of mediation sessions indicate a mix of facilitative, education, and more directive styles of mediation. Outcomes of CharlottesvUle Custody Mediation Project. The Charlottesville research investigated a broad array of outcomes. Settlernent rates. In the mediation group, 77 percent ofthe thirty-five families reached either written or verbal agreement regarding their child issues. Only four ofthe eight families not reaching agreement proceeded to court. Among the litigation group, 72 percent appeared for hearings in domestic relations court for judicial determination. Satisfaction of participants. Mothers reported more satisfaction on a number of measures assessing both process and outcomes than men in both the mediation and litigation groups. Compared to fathers who litigated, fathers who mediated were substantially more satisfied on all nineteen measures, including aspects of the process, its effects on them, their relationship with their children, and their relationship with their former spouses. Fathers in the litigation group were significantly less likely to feel

12

KELLY

their rights were protected compared to fathers in the mediation group or either group of women. In comparison with mothers who litigated, mothers who mediated were less satisfied with the process, including the fairness of decisions, but they did not differ on feeling their rights had been protected. They were more satisfied than mothers who litigated on several other measures, including feeling understood, concern shown for self and for children, and whether problems were caused with their spouse (Emery, Matthews, and Wyer, 1991). At a one-year follow-up, mothers who mediated were less satisfied with the process and outcome of dispute resolution than they had been previously, but attrition of satisfied mothers in the follow-up sample may have contributed to this change (Emery, Matthews, and Kitzmann, 1994). Women in both groups remained more satisfied than fathers on a number of other measures of outcome, and "the majority of the study's findings point to the disadvantage of men in litigation, not to the disadvantage of women in mediation" (Emery, 1994, p. 187). Mediation parents more often reported that they each had won "somewhat" of what they wanted, compared to litigation parents, where mothers felt they had won "quite a bit" of what they wanted and fathers had won only "a little" (Emery, 1994), providing empirical support for the claim that mediation is more likely to produce modest win-win feelings in both parents rather than the widely divergent win-lose feelings characteristic of adversarial processes. Gender differences in satisfaction, or lack of them as in the California custody and comprehensive mediation studies (Center for Families, Children and Courts, 1993a; Kelly, 1989), must be set within the context of the legal and social climate of the research setting. California adopted gender-neutral statutes regarding custody much earlier than did Virginia, and permitted joint legal and physical custody options earlier than most other states. Sole maternal legal and physical custody was the most common outcome in Virginia at the time this study was conducted (Emery, 1994). In contrast, by 1990 in California, approximately 85 percent of parents had joint legal custody and reached access agreements (or had orders imposed) that reflected a broader array of access or shared physical custody arrangements. Changes in conflict and parent-child relationships. One of the most striking findings of this project was that fathers who mediated remained more involved with their children one year and twelve years later, compared to fathers in the litigation group (Emery, 1994; Emery and others, 2001). There were modest reductions in parent conflict in the mediation

Family Mediation Research

13

group that were sustained over the first postresolution year as well. Mediation did not differentially improve adult mental health compared to the litigation group. Efficiencies in time. Mediation settlements were reached in half of the time required by the litigation group, and significantly reduced the number of hearings needed in the court (Emery, Matthews, and Wyer, 1991). Outcome patterns. Whereas sole maternal physical custody was the predominant outcome for both groups, joint legal custody was more often an outcome in the mediation sample compared to the litigation group (Emery, Matthews, and Wyer, 1991). Actual parenting time was not substantively different in the two groups, including the number of days with the nonresidential parent. As in other studies (Ellis and Stuckless, 1996; Kelly, 1993), mediated agreements were more detailed as to days and times of transitions, other child-related details, and child support, whereas litigated agreements contained more vague language (for example, "reasonable visitation to the father"). Compliance. Fathers who mediated complied more often with child support orders than did fathers who litigated (Emery, Matthews, and Kitzmann, 1994).
Access and Visitation Grant Mediation Programs

The U.S. Department of Health and Human Services (2002) funded studies infivestates that provide access mediation to the IV-D population, with the objective of determining if mediation would facilitate and increase access rights for noncustodial parents and increase visitation. Findings were from 254 cases: 190 cases combined from four states (Nevada, Connecticut, Oklahoma, and Illinois), and 64 cases, analyzed separately because of programmatic differences, in Ceorgia. Structural Elements of the Access and Visitation Mediation Programs. Randomly selected cases from nine community and court-based programs offering mediation for access problems in 2001 were examined. Participants were mothers and fathers in the IV-D program, a child support agency funded for collecting and enforcing child support orders. Mediation and court files and child support payment records were reviewed, and a telephone survey was conducted for 125 parents. Information was not provided on the timing ofthe intervention, or on the training, qualifications, or orientation ofthe mediators. Services were offered free of charge, and the average mediation session was 1.4 hours. In the majority of programs, referrals

14

KELLY

came from child support offices and the courts, and attendance was voluntary; three programs had mandatory participation. Outcomes of DHHS Access and Visitation Program Research. Both settlement rates and satisfaction were studied in the DHHS research. Settlement rates. In 76 percent of cases, mediation facilitated noncustodial (NCP) parents' access rights through an agreement on visitation plans. In 69 percent of cases, NCPs had no prior access rights, and of these, 77 percent gained access in a mediated visitation agreement. Access rights increased from zero to nineteen hours, on average, for standard visitation and from zero to thirteen days for vacations and holidays. Of the 31 percent of NCPs with prior access rights, 74 percent successfully reached a visitation agreement in mediation with the custodial parent (CP), and in this group, there was slightly increased access (U.S. Department of Health and Human Services, 2002). Satisfaction ofparticipants. Parents were asked to what extent they would recommend or advise against the mediation program. Eighty-one percent of CPs would strongly or somewhat recommend, with no difference between those who did and did not reach agreement. Among NCPs, 71 percent of NCPs would recommend, with 74 percent of those who reached agreement versus 59 percent who did not reach agreement indicating they would recommend mediation. Outcome patterns. In follow-up telephone surveys of one hundred cases, 42 percent of parents who reached an agreement reported an overall increase in noncustodial parents' visits following mediation. Gender differences were found, with 39 percent of CPs and 53 percent of NCPs reporting that visits increased. For 33 percent, visits stayed the same, most often those parents who reached no agreement, and 11 percent reported a decrease in visitations. In the six months after reaching a mediated agreement, the majority of both CPs and NCPs reported that more weekly visits occurred compared to the six months prior to mediation; in addition, visits were more regularly scheduled and were cancelled less often. One-third of CPs and two-thirds of NCPs who reported increased visits reported an improvement in their child's behavior after mediation. Based on a subset of 111 cases from four states, 61 percent of NCPs increased the percentage of current child support they paid after mediation and 27 percent decreased the percentage. Prior to mediation, NCPs paid 52 percent of what was owed in child support; after mediation, NCPs

Family Mediation Research

15

paid 70 percent of what was owed, significantly above what the national child support system collects. Comprehensive Divorce Mediation Projects Far fewer studies have assessed comprehensive divorce mediation encompassing division of property, spousal and child support, and child custody and access issues. Comprehensive divorce mediation is most often offered in the private sector on a fee-for-service basis by lawyers and mental health professionals. The two studies considered here, the Divorce and Mediation Project in California (Kelly, 1989, 1990, 1991a, 1991b, 1993) and the Divorce Mediation Pilot Study in Ontario, Canada (Ellis, 1994; Ellis and Stuckless, 1996), represent a private and a public sector project.
California Divorce and Mediation Project

The Divorce and Mediation Project was a multidimensional study of the comparative legal, economic, psychological, and relationship effects of mediation and adversarial divorce on the participants during and after the divorce process. The longitudinal study collected data at five points in time, where time 1 was the beginning of divorce, time 3 final divorce, and time 5 two years after the final divorce. At time 1, the mediation sample consisted of 212 individuals (106 couples), and the adversarial divorce group consisted of 225 respondents (including 47 couples). The overall sample of 437 was mostly well educated, primarily white, and middle class. The mediation sample had more education and minor age children compared to the adversarial group, but median income did not differ, and there were no differences in amount of marital conflict or anger at spouse as reported at the beginning of divorce. Structural Elements in the Divorce and Mediation Project. This study compared process and outcomes in the private sector for two groups: those who voluntarily chose to mediate their divorces from 1983 through 1985 at a nonprofit mediation center and agreed to participate in the study, and those who filed a divorce petition between 1984 and 1986, were randomly selected from court records, agreed to participate in the study, and had an attorney representing them in divorce proceedings. For the mediation couples, the timing ofthe intervention varied, as one-third of couples had not yet separated at entry into mediation and the remaining couples began mediation several months to more than a year after separation.

16

KELLY

Comediations were conducted by a lawyer and two psychologists specifically trained in mediating property, support, and custody issues and with knowledge of family law. The mediation was a task-focused, facilitative, and problem-solving model, and responsibility for decision making remained with the parties on all issues. Final agreements were recorded in detailed memoranda of understanding, and any issues not resolved in mediation were left for subsequent attorney negotiations or court hearings. The average number of mediation sessions for those reaching agreement on all issues was ten (approximately fifteen hours), which varied with the legal and psychological complexity of issues, amount of conflict, and motivation of clients. Sliding-scale fees ranged from $40 to $ 120 per hour per couple, with an average fee per couple of $110 per hour, which was comparable to family law attorney fees at that time. Mediation respondents were encouraged to consult attorneys as needed and used counsel to prepare the legally binding marital settlement agreement and to file divorce papers. The mediation clients had high levels of self-determination with respect to selecting and ordering issues for discussion and negotiation, length of sessions, the pace ofthe mediation, the use of outside counsel and additional experts, and all temporary and final decisions. Outcomes of California Divorce and Mediation Project. This comparative research provides insights on a variety of outcome measures. Settlement rates. Four groups were distinguished within the mediation sample. "Comprehensive completers" (50 percent) reached agreement on all issues pertaining to their divorces, resulting in a memorandum of understanding. They were not more cooperative than the other subgroups, but both spouses reported similar levels of knowledge about finances. A second group, "partial completers" (8 percent), were able to reach full written agreement within a particular area, usually custody and parenting or child support, but not reach full resolution on the other issues. This group was more likely to have attorneys active on their behalf prior to coming to mediation and had the largest discrepancy between the husbands' and wives' self-ratings of financial knowledge. Thus, the overall rate of reaching partial and full agreement was 58 percent. Among those who terminated mediation were two groups: the "productive terminators" (15 percent) resolved one or more critical issues, most often related to details of their separation including visitation or temporary support, but did not return to negotiate final divorce issues. This group scored highest on "interest in

Family Mediation Research

17

reconciliation," and a number later reached agreement on their own. The "true terminators" (26 percent) were unable to reach agreement on anything, and as a group they were more likely to report that they had an angry, demeaning spouse or an emotionally unstable, substance-abusing spouse. They also gave the lowest ratings of their spouses' honesty, fairmindedness, and level of cooperation. These settlement rates are somewhat lower than rates reported in the court-based custody mediation studies in Virginia and Colorado but comparable to a large Canadian study of courtbased comprehensive divorce mediation, which found full agreement in 49 percent of cases, partial agreement on 15 percent, and 64 percent when combined (Richardson, 1988). Comprehensive divorce mediation involves settling multiple complex issues, which requires an ability to comprehend all the issues and remain in mediation for the longer course. Satisfaction with mediation. Significant group differences were found on eighteen of forty items from a fifty-four-item Client Assessment of Mediation Services (CAMS) scale assessing participants' perceptions of and satisfaction with their respective divorce processes (see Kelly and Cigy, 1988; Kelly, 1989). Between 65 and 82 percent of all respondents viewed their mediators and attorneys as warm, sympathetic, and sensitive to feelings; helpful in standing up for their rights in disagreements with spouses; staying focused on the important issues; and having clear and sufficient information for decision making, with no group differences. The mediation group rated their mediators as more skillfiil and more helpful in proposing ways to resolve disagreements and getting to workable compromises, compared to the litigation group ratings of their attorneys. Mediation clients, particularly women, viewed mediation as more empowering than did the adversarial men and women in helping them assume greater responsibility in managing their financial affairs, and in better understanding their spouses' points of view. Men in both groups rated their attorneys or mediators as more often favoring their spouses' point of view than did the women, but there were no sex differences within just the mediation group. Seventysix percent of mediation women and 62 percent ofthe men indicated that mediation helped them to become more reasonable with each other, compared to 26 and 39 percent ofthe adversarial men and women, respectively. More than half of the adversarial group reported that the divorce process had caused a deterioration in their communication compared to 11 percent ofthe mediation group. Finally, on a global measure of satisfaction, 69 percent of mediation respondents were somewhat to very satisfied, compared to 47 percent of adversarial men and women, with no sex differences.

18

KELLY

Changes in conflict and coparental relationships. Mediation respon-

dents with minor children reported less conflict during the divorce process, at final divorce, and in the first year after final divorce on a number of measures compared to parents in the adversarial group. Since there were no group differences at time 1 on these measures, it appeared that mediation had the effect of diffusing and reducing interparental conflict, compared to the adversarial process, at least for a period of more than two years. By two years postdivorce, there were no longer any group differences in frequency of conflict, as the adversarial group gradually reduced their conflict. However, when conflict occurred, the mediation parents used a more direct and mutual style of resolving their conflict, compared to the adversarial parents, who most often avoided each other (Kelly, 1991a). Compared to the adversarial process, mediation was more effective in increasing the general level of cooperation between beginning of divorce and final divorce, after controlling for initial baseline differences. These differences were still evident at the end ofthe first year of divorce but were no longer significant two years postdivorce. However, mediation parents continued at two years postdivorce to seek parenting help from each other more often than the adversarial parents, were more likely to accommodate any requested changes in parenting schedules, and could communicate by telephone with their children whenever they or the children wanted to compared to adversarial parents (Kelly, 1991a). These findings may be related to the fact that at final divorce, the mediation group indicated more beneficial effects on their ability to be reasonable with each other and that communication improved somewhat compared to the adversarial respondents. They also perceived their former spouses as less angry, even though objective measures of anger of former spouse did not significantly differ (Kelly, 1989). Divorce mediation was not more beneficial than the adversarial process on a number of standardized and objective measures of psychological adjustment. Instead, the passage of time was associated with a reduction of symptoms, such as depression, anxiety, and paranoia, for both groups at final divorce, and at one and two years postdivorce (Kelly, 1990, 1991a). Furthermore, no changes in children's adjustment were associated with either dispute process. Similar findings were reported elsewhere (Emery, 1994; Walker, McCarthy, and Timms, 1994). The claims that mediation would improve psychological adjustment were quite unrealistic, particularly given the specific dispute resolution focus of mediation and the brevity ofthe intervention.

Family Mediation Research

19

Efficiencies in time. No differences were evident in the length of time it took mediation and adversarial group respondents to obtain their final divorce. There were, however, significant links between marital and psychological variables and the number of hours required to complete mediation and the costs associated with adversarial divorce. Outcome patterns. In mediation, parents agreed to joint legal custody more often than did the adversarial parent group, as was reported in other studies (Emery, 1994; Pearson, 1991; Richardson, 1988). Mediation also resulted in parenting plans that were more generous in parenting time with children for the noncustodial parent (Kelly, 1993). Mediation parents more often negotiated fathers' time with children in the range of 20 to 39 percent, compared to the adversarial parent outcomes, in which fathers more often had time with their children between 0 and 20 percent (every other weekend or less). These findings of more time allocated between fathers and children in the mediation sample are in contrast to Emery's custody mediation study (1994), where contact with children every other weekend was most common in both groups. Mediators in that project discouraged parents from negotiating more extended contacts or shared physical custody parenting when there was high conflict or other difficult issues, which may have contributed to the differences (Emery, 1994). In the California study, mediators did not directly caution parents who were discussing higher amounts of parenting time with fathers, but focused on creating confiict-free transitions of children and parental communication regarding their children. The agreements of mediation parents contained much greater specificity about decision making regarding children, parenting plan details, and parental responsibilities for health insurance and unreimbursed medical costs. No differences were found in child support between mediation and adversarial groups, but the mediation fathers did pay for more "extras" for their children, including lessons and extracurricular activities, and were more likely to make some agreements regarding their children's college expenses and support beyond age eighteen than did the adversarial group. More recent research suggests that such extras may be associated with fathers' having more time with their children rather than an effect of mediation (Fabricius and Braver, 2003). No analysis of property division was conducted because ofthe complexity of analysis required and some expectation that the fifiy-fifiy community property division laws might not result in differences between the two processes (Pearson, 1991). The mediation clients rated their division of property as fairer than did the adversarial respondents (Kelly, 1990).

20

KELLY

Cost savings. The combined couple costs for the adversarial group were 134 percent higher than for the mediation group, despite comparable hourly fees for both groups. For the mediation group, total costs included all mediation hourly fees (the mean was $2,224), and external consulting attorney, accountant, and appraisal fees associated with the divorce for each spouse, for a mean total of $5,234 (the median was $3,428). The mean total costs for couples in the adversarial sample were $12,226 (for men, $6,850; for women, $5,376). The two groups did not differ on a divorce issues complexity scale or in the extent of marital conflict, initial level of anger at spouses, cooperation at beginning of divorce, amount of anticipated disagreement about issues to be resolved, or household income (Kelly, 1991). Thus, there is strong support for assuming that the differential cost of divorce was the result ofthe mediation process itself The inclusion of marital and psychological variables in this research provided a unique opportunity to explore divorce costs related to other dimensions. When spouses in mediation reported that they divorced primarily because of a high-conflict or demeaning relationship or because of a substance-abusing or emotionally unstable spouse, mediation took longer and therefore cost more (see Cigy and Kelly, 1992, for analyses of reasons for divorce). Similarly, poorer-quality marital communication, poorer cooperation, and perceptions of the spouse as dishonest, or taking advantage of the respondent, or lacking in fair-mindedness were all associated with longer and more costly mediations, although these partners did manage to complete the process. In contrast, mediation clients who rated their spouses as "good spouses" (a three-item scale measuring flexibility, fairmindedness, and ability to compromise) had lower mediation costs. Interestingly, neither depression nor level of anger toward the spouse correlated with the time required to complete mediation (Kelly, 1990). Among the adversarial sample, attorney fees were higher for the nien when they rated their spouses as dishonest and felt taken advantage of by them. For women, higher legal fees were associated with being angrier at their spouses, perceiving their spouses to be dishonest, lacking in fair-mindedness, and to have taken advantage of them during the marriage. In the adversarial group, neither level of cooperation nor reason for divorce was related to attorney fees. Compliance. The majority of respondents in both groups were complying with the terms of their final divorce orders, perhaps consistent with the relatively advantaged socioeconomic status of participants. But there was significantly more compliance immediately after divorce in the

Family Mediation Research

21

mediation group on an eleven-item compliance scale, compared to the adversarial sample, and more compliance on the separate items of paying spousal support, final division of personal property and household furnishings, and division of community property. These differences were still evident at one year postdivorce but disappeared at two years postdivorce, in part because the adversarial group become more compliant over time in child support payments. Women in both groups reported less compliance with child support orders than did men. Noncompliance in the mediation group was primarily late payments, whereas in the adversarial group, noncompliance was nonpayment or paying less than ordered. Many variables expected to contribute to noncompliance at two years postdivorce did not, including satisfaction with the divorce outcome, divorce process group, conflict in the marriage, and the amount of anger at spouse at final divorce and two years later.
Family Mediation Pilot Project-Ontario, Canada

This study focused on 361 men and women who were means tested and approved to receive a legal aid certificate for immediate assistance with separation and divorce, and who selected either a lawyer negotiation or mediation path for resolution of their disputed issues. Canadian lawyers are required by law to describe mediation as one option for resolving marital matters. Structural Elements ofthe Eamily Mediation Pilot Project. Participants in this court-connected mediation study were voluntary and self-selected into either a mediation or adversarial sample on the basis ofthe information provided by lawyers and on referral from the court. Spouses indicating an interest in using mediation had the opportunity to attend a separate family law information meeting prior to beginning mediation, followed by an individual interview and screening for abuse and large power imbalances to confirm the choice and feasibility of mediation. Some settings used open mediation, involving recommendation to the court, and others were closed mediations, that is, confidential. Clients who selected lawyer negotiations were seen in the attorney's private practice. Services for both groups were offered free of charge. The Hamilton Unified Family Court mediation settings offered comprehensive mediation for custody and access, support, and property division. More than two-thirds of clients in both samples indicated that access (visiting) was the major issue to be settled, 40 percent wanted the issues of

22

KELLY

custody and child support settled, and 10 percent wanted spousal support settled. The timing of both interventions was from five to twelve months after separation disputes were apparent, during which time some participants initiated motions and other legal action. Clients who chose lawyer negotiations were poorer, less well educated, had been separated for a longer period of time, and had been more recently abused by their partners, compared with those who chose mediation. Women were also more likely to choose lawyer negotiations than were men (Ellis, 1996). Mediators were trained and experienced in family law mediations. Outcomes of Eamily Mediation Pilot Project. The Family Mediation Pilot Project studied standard outcomes like settlement rates and satisfaction, but also looked at spousal violence rates. Settlement rates. Along the four major issues of access, custody, child support, and property division, settlement rates in the mediation sample ranged from 40 to 80 percent, with an average rate of 60 percent, compared to approximately 80 percent in the lawyer negotiation group. Settlement rates were higher when mediation occurred prior to parties' being involved in any legal proceedings. Satisfaction with process. Among the 169 mediation clients, 66 percent indicated they participated in a process they judged to be the best way for them, in contrast to 48 percent ofthe 192 lawyer clients who thought that process was the best way. Far more lawyer than mediation clients indicated they had participated in a process they believed was the worst way for them to settle issues of separation and divorce (28 versus 1 percent). Clients who reached agreement in both processes were more satisfied than those who did not, and mediation clients were more satisfied than those participating in adversarial processes (Kelly, 1989). On each separate issue in contention, the percentage of satisfied mediation clients was higher than the settlement rate percentage. In contrast, in the lawyer negotiation group, satisfaction percentages were lower than settlement rates in each area, with the exception of custody, where women in the lawyer group were more satisfied than the men, and the men and women in the mediation group. These women were most likely to petition for and receive sole custody through negotiations, hearings, or trial compared to the mediation group. The least satisfied of all groups were men in the adversarial group, also reported elsewhere (Emery, 1994; Kelly, 1989). In the Canadian study, when lawyers participated in mediation sessions as comediators (and gave advice), mediation clients were less satisfied than were clients where no lawyers (or judges) were involved.

Family Mediation Research

23

Spousal violence. Three forms of abuse (emotional, verbal, and physical) were studied pre- and postprocessing for both groups. On a four-item composite abuse scale (also including intentional hurting), a greater number of mediation clients reported being physically, emotionally, and verbally abused prior to the mediation than did lawyer clients. However, the differences were no longer significant postintervention, suggesting that mediation made a greater contribution to reducing the violence than did lawyer processing. The use of affidavits (declarations) in the lawyer process that were personally attacking and hurtflil rather than fact based accounted in part for the continuing violence in the lawyer group (Ellis and Stuckless, 1992). The most important factor responsible for decreases in abuse postintervention was physical separation, which decreased opportunities for both conflict-instigated and control-instigated abuse in both groups. All three forms of abuse decreased over time following the lawyer and mediation processes, with verbal and emotional abuse decreasing more for mediation clients in the first twelve months (Ellis and Stuckless, 1996). These researchers (1992) also found more postprocess violence in a onesession coerced mediation provided by Legal Aid when compared to courtconnected voluntary mediation (publicly funded) averaging six sessions. Outcome patterns. Women in the mediation sample were more likely to get (and be satisfied with) the child support amount they wanted compared to the adversarial women. A greater proportion of adversarial group women got the sole custody they wanted compared to the mediation women (85 versus 67 percent). More joint legal custody agreements were reached in the mediation group than in the lawyer negotiation group, similar to Emery (1994) and Kelly (1990, 1993). In the mediation sample, abused and nonabused wives were equally likely to say that they were equal to or better able than their partners to stand up for themselves and state their positions (Ellis and Stuckless, 1996). Differences between husbands and wives in the mediation group in income and various measures of marital power were not significant predictors of custody, access, support, or property division outcomes. Compliance. Compliance problems were reported in greater numbers in the adversarial compared to the mediation group, and within the mediation group, those participating in one-session mediations reported more compliance problems than did those in multiple sessions. Wife abuse during the six months prior to separation was a significant predictor of compliance problems reported by female mediation clients. As in the California study, the presence of children was associated with more compliance problems.

24

KELLY

Child Protection Mediation

In response to an increase in child maltreatment filings in the 1980s and the recognition that traditional adversarial processes had serious limitations in settling these cases, a number of courts initiated programs that offered mediation in the dependency courts. The earliest program was started in 1983 in Los Angeles Juvenile Court, and twenty-one programs were operating in California in 2002 (Center for Families, Children and Courts, 2003b). Three research programs in California, Colorado, and Ohio are described.
Child Protection Mediation in Five California Courts

Five pilot child protection mediation programs were given the opportunity to customize their services based on their own needs and caseloads. Four programs used mediation at the first major legal step: whether the child in out-of-home placement should be placed under the authority ofthe court. Mediation occurred at the next major step of case disposition in nearly all the counties, when a plan was developed to indicate which programs and services parents would need to attend to aid in reunification with the child. Most often, both the petition and the treatment plan were mediated in the same session. Several courts also used mediation at later stages in the cases, including at the permanency planning hearing, and dealt with noncompliance and problems with the placements (Thoennes, 1997). Judges had discretion to refer cases to mediation, typically when jurisdiction and disposition were contested, and parent participation was mandatory with such referrals. Additional participants often included a representative of the child protection agency, the parents' attorneys, or other family caretakers. A confidential facilitative mediation model, free for parents, was provided by trained and experienced mediators in teams or solo, generally for more than ninety minutes. Data were collected for 606 mediated cases and 223 nonmediated cases.
Outcomes of California Child Protection Mediation Project Evaluation

Outcomes in child protection mediation are similar to those studied in other forms of family mediation. Settlement rates. Based on mediator ratings, between 60 and 80 percent of cases reached full agreements, and 90 percent reached some form of agreement. Cases were more likely to settle at the disposition or

Family Mediation Research

25

postdisposition stages rather than earlier. These families had serious problems: many had been in the child protection system before, half had a known drug problem, and one-third had been the subject of a criminal court filing (Thoennes, 1997). Satisfaction. More than 90 percent of parents felt that they had been understood in mediation, had a chance to talk about important issues, and felt mediation clarified what they needed to do in order for the agency to close their case. Most parents felt mediation was better than a court hearing. Initial resistance among all professionals to participating in mediation was short-lived. Outcome patterns. Mediated agreements were more detailed on specifics of visitation arrangements and more generous, and they included more specific services to be provided to the child when compared to nonmediated agreements. In addition, communication problems between parents and caseworkers were more often addressed, and parents were more likely to acknowledge that they needed to cooperate with the treatment plan and needed help in changing behaviors. Cost savings. Mediation of child protection cases in the long run appeared to help courts avoid repeated hearings on the same case. At six months postdisposition, based on data from four sites, 88 percent of the mediated cases did not require a contested review compared to 53 percent of the control cases. Many professional participants cautioned that while mediation may avoid repeat hearings and three- to four-day trials, appropriate mediation services take resources and commitment to making it work. Compliance. Combined site data indicated that there was complete compliance for 42 percent of the mediation cases compared to 25 percent ofthe comparison group at six-months postdisposition.
Child Protection Mediation in the Colorado Fourth Judicial District

Information from 243 child protection mediations completed between 1997 and 1999 and 49 cases from court records of comparable cases without mediation that had court action between 1997 and 1998 was collected to assess the effectiveness of mediated cases and participants' reactions (Thoennes, 2000). Structural Elements in the Colorado Program. Mediated cases were evenly divided between predisposition and postdisposition cases and between families that been the subject of previous dependency filings and those that were in the court system for the first time. All cases in which a contested

26

K-ELLY

hearing or trial was requested were automatically ordered into mediation by judges. The most common types of maltreatment alleged were neglect (61 percent) and physical abuse (29 percent), and the majority of parents were rated as having major problems such as substance abuse or mental disabilities. The average mediation was 1.8 hours (the range was one to five sessions), with most cases resolved in a single session (Thoennes, 2000). Mediators were experienced professionals who viewed their role as a facilitator, not an evaluator or decision maker. The average number of participants in the mediation was 6.5, with a mean of 4.3 professionals in attendance. Outcomes of Colorado Eourth Judicial District Child Protection Mediation. This research reports positive effects of child protection mediation. Settlement rates. Nearly 70 percent of the mediation cases reached consensus about all issues pending in the case, 20 percent reached partial agreement, and 11 percent reached no agreement in mediation. Settlement rates did not differ based on type of maltreatment, the stage at which the case was mediated, or the issues being discussed. Satisfaction. Initial reservations of professional participants about mediating these cases were significantly diminished, with professionals indicating they would not want to return to adversarial processes for handling these cases. Mediation provided a critical time in a child protection case when everyone was present and prepared to discuss the case. Cost efficiencies. Using conservative cost estimates of avoidance of trials, expert witnesses and repeated psychological evaluations, and trial preparation time for attorneys and caseworkers, analyses projected that the child protection mediations allowed the Fourth Judicial District to reduce its cost per case by 13 percent.
Hamilton County Permanent Custody Mediation Program

The Juvenile Court of Hamilton County (Cincinnati, Ohio) received U.S. Department of Health and Human Services funding to provide and evaluate mediation services to families that were the subject of permanent custody filings. Cases randomly assigned to the mediation and control groups resulted in forty-nine cases with a permanent custody mediation and thirty-seven comparable cases assigned to the control group. The focus of the study was whether mediation would shorten the time between filing and the termination hearing, whether biological parents in mediation

Family Mediation Research

27

would be more involved in the permanency planning process, and whether mediated termination agreements would include more provisions for information sharing or some contact between adoptive parents and biological parents. Structural Elements in the Permanent Custody Mediation Program. This program compared mediated and contested cases in the last and most difficult stage of child protection cases, that of a filing for termination of parental rights. Parent participation was mandated, and mediations usually included the mother, mother's attorney, caseworker, attorney for the agency, and the child's advocate. Fathers attended in 38 percent of mediated cases. Mediators used a facilitative model of confidential mediation and emphasized that parents had the right to choose not to make an agreement and the right to go to court. Parents were informed about binding and nonbinding agreements and the permanency of termination of parental rights. If an agreement was reached, parents would meet immediately with the magistrate and have the stipulated agreement entered into the court record if it was determined that they understood the terms of the agreement, had entered into it voluntarily, and alternatively could go to court. Mediators were required to have fifty-two hours of mediation training and a four-day training specific to child protection lavv^s, roles and practices, types of cases, and violence, and comediated with experienced dependency mediators. Outcomes ofthe Permanent Custody Mediation Program. As with other forms of family mediation, the results of this research confirm mediation efficacy. Settlement rates. Nearly 40 percent of mediated cases reached an agreement about permanent custody, which is remarkable given the emotionally difficult issue in dispute. While settlement did not appear to be influenced by the type of maltreatment, a prior history of child protection reports, or mental or physical health issues for the children, the length of time a child had been in out-of-home placement was related to settlement, with a settlement rate of 57 percent among those with children in placement for more than a year versus 31 percent for those in placement for less than a year. The settlement rate was also higher in families where just one child was the subject ofthe permanent custody filing (56 percent), as contrasted to a 30 percent settlement rate when two or more children were

28

KELLY

involved (Thoennes, 2002b). In approximately half of cases mediating agreement that discussed open adoption, some form of continued visiting or periodic exchanges of written information was agreed on, the former more often with children permanently placed with relatives. Satisfaction ofparticipants. Most parents expressed satisfaction with aspects of the process, saying they felt heard and understood (74 percent), were given a chance to talk about what they really wanted (87 percent), and were treated with respect (83 percent). Overall, nearly 70 percent of parents felt mediation was better than going to court, with satisfaction higher among the setdement group. Despite disapproval of mediation at the beginning, rates of satisfaction among professionals ranged from 70 to 100 percent (Thoennes, 2002b). Efficiencies in time and money. About 70 percent of cases reaching agreement proceeded directly to the court to enter the agreement on the record. Using an analysis of estimated costs per case in mediation and the comparison sample, the resolution by mediation of a permanent custody case was estimated to cost 39 percent less than processing a case in the absence of mediation (Thoennes, 2002b).

Summary of Findings
Using a variety of methodologies, measures, and samples, the nine studies described suggest strong support for the use of mediation in family disputes. In public and private sectors, in voluntary and mandatory services, and when provided both early and late in the natural course of these disputes, family mediation has been consistently successful in resolving custody and access disputes, comprehensive divorce disputes, and child protection disputes. Mediation has given evidence of its power to settle complex, highly emotional disputes and reach agreements that are generally durable. Settlement rates in custody, comprehensive divorce, and child protection mediations generally range between 50 and 90 percent, with the exception ofthe most difficult cases involving parental termination. Mediation appears to work with angry clients and sometimes for those with serious psychological and family problems. What is required are experienced and trained mediators. However, profound distrust and a lack of fairmindedness on the part of one or both partners more often interfere with reaching agreements (as is true in litigation processes as well).

Family Mediation Research

29

Client satisfaction has been surprisingly high in all studies and settings on a large number of process and outcome measures. As would be expected, satisfaction is higher when clients reach agreement as opposed to no agreement, and those who use custody mediation are substantially more satisfied than parents using other court processes. Repeatedly, clients indicated that they felt heard, respected, given a chance to say what is important, not pressured to reach agreements, helped to work together as parents, and felt their agreements would be good for their children. Mediation clients in the private sector are significantly more satisfied on almost all measures of process and outcome than are those using adversarial divorce processes. Where gender differences in satisfaction were found, the legal context appears to be an important factor, as are the issues in dispute. While mediation does not improve psychological adjustment in measurable ways, there are other important benefits for all family members. When contrasted to parents in adversarial processes, parents using a more extended mediation process experience a decrease in conflict during divorce, and in the first year or two following divorce, they are more cooperative and supportive of each other as parents and communicate more regarding their children, after controlling for any preintervention group differences. One astonishing result has been that twelve years following divorce, fathers in mediation remained more involved with their children compared to the litigation fathers. Cautions as well emerge from the literature. Consistently, 15 to 20 percent of parents of both sexes are dissatisfied with aspects of mediation process and outcomes. Although this represents half the rate of dissatisfaction of adversarial clients, it is important to know if this reflects a more rushed or coercive mediation process, untrained or inept mediators, or parents who are angry and dissatisfied with any divorce process and outcome that does not produce what they expected or wanted. With the trend to limit court custody mediation to one session, more difficult cases with multiple serious issues most likely will not be given sufficient opportunity to settle, and settlement rates may decline. Future Directions in Family Mediation Research Family mediation research has been elucidating and promising, and has generated far more understanding than the scant and mostly superficial assessments of various adversarial divorce processes, arbitration, or collaborative law. To date, however, this research has not led to more complex

30

KELLY

second-generation research, in part due to a chronic lack of research funding for mediation, the complexity of what is required, and an apparent diminishing interest in research questions in the field. Among the issues that need investigation are the relative merits of different mediation models used in family mediation, including facilitative, problem-solving, transformative, and more evaluative models of mediation (for descriptions of these models, see Folberg, Milne, and Salem, 2004). No empirical research exists that explores and compares the efficacy of these different models, the purity of their processes and practices,. for whom and what type of disputes they are most effective, and their respective outcomes. This. is surprising given the ongoing and somewhat contentious debate regarding these models. No research has yet investigated the relative effectiveness of mediation models that work with couples conjointly compared to models that use separate sessions exclusively. Are there different outcomes in client conflict, cooperation, and postdivorce communication for parents? Newer hybrid models, such as parenting coordination, which combine mediation, arbitration, and education for highconflict parents, have also received little research attention (Kelly, 2002). The exception has been research conducted on specialized strategies and mediation models developed for high-conflict and violent parents who continue to chronically litigate child issues after the divorce (Johnston, 1994; Johnston and Campbell, 1988; Kelly, 2002; Smyth and Moloney, 2003; Pruett and Johnston, 2004). These more intensive therapeutic mediation models, typically longer in length, integrate individual and group mediation sessions with counseling and have been demonstrated to be effective with this difficult population. A related issue is what type of mediation models and services are necessary and most effective for participants when there is a history of domestic violence, particularly since a sizable number of both women and men enter, and sometimes prefer, mediation to other adversarial alternatives (Ellis and Stuckless, 1996; Newmark, Harrell, and Salem, 1995). Since many states quickly passed legislation prohibiting custody mediation where domestic violence had occurred or was alleged, flirther study was discouraged. These statutes and policies were based on the belief that domestic violence was primarily that of males battering women. Recent research documents other less severe, nonescalating, and more commonly occurring categories of violence in relationships, including bidirectional or common couple violence, female violence, and separation-engendered

Family Mediation Research

31

violence, each with its distinctive features and histories. Among these partners are many who are not afraid of each other, cease their violence at separation, and are capable of mediating (see Johnson and Ferraro, 2000; Johnston and Campbell, 1993; Statistics Canada, 2001). Research in mediation process issues remains quite limited, in part because such research is complex, expensive, and time-consuming (Kelly, 1996). The field would be advanced considerably if analyses were conducted of mediator behaviors and interventions, participant characteristics and behaviors, and the relationship of these to outcomes. As an example of the latter, couples with high mean dyad levels of anger did less problem solving, as did those with higher mean contentious behaviors, compared to those with disparate levels of anger. Each of these was among the best predictors of failure to reach agreement (Bickerdike and Littlefield, 2000). Such research has implications for mediation process and the active use of mediator techniques such as conflict management and structuring behaviors. Understanding the interaction of emotions and personality attributes that individuals bring to the mediation setting, and the styles and behaviors of mediators that diminish or enhance the likelihood of reaching agreements, would help the field define and refine practices, improve effectiveness, and promote excellence in the field.
Notes 1. Research was conducted under the auspices of the Administrative Offices of the Courts, Judicial Council of California. See www.courtinfo.ca.gov/programs/ cfcc/resources/publications/articles. 2. Relitigation rates in child custody matters do not necessarily represent a failure or breakdown ofthe mediation agreement, or, for that matter, a failure of attorneynegotiated or judicially imposed orders. Relitigation may serve legitimate purposes and point to breakdowns in agreements among contentious parents. Some parents must relitigate when a parent relocates to modify future parenting arrangements, others return to renegotiate a more appropriate parenting plan as children's developmental needs change, and still others return when parents are not meeting their parenting responsibilities or are interfering with access. In Colorado, in the first two years following final divorce orders, 38 percent of the comparison group returned to court compared to 24 percent of the mediation group. On three specific issues relitigated (parenting time modifications, child support contempt, and other modifications), significant differences again favored the mediation group (Thoennes, 2002). In the Charlottesville project, relitigation rates were quite high in both groups during the first two years (nearly two-thirds returned to court). In contrast, in the California divorce and mediation project, at one and two years

32

KELLY

postdivorce, relitigation rates were much lower in both groups (15 percent of all respondents and 20 percent among the parent subsample). In the Canadian study, adversarial clients reported higher rates of relitigation during the first two years postdivorce, but byfiveyears aft:er divorce, there were no differences, as 10 percent in both groups returned to court (Ellis, 1996). References Beck, C.J.A., and Sales, B. D. "A Critical Reappraisal of Divorce Mediation Research and Policy." Psychology, Public Policy, and Law, 2000, 6{A), 989-1056. Bickerdike, A. J., and Litdefield, L. "Divorce Adjustment and Mediation: Theoretically Grounded Process Research." Mediation Quarterly, 2000, 18 (2), 181-201. Center for Families, Children and Courts. Client Evaluations of Mediation Services: Perspectives of Mothers and Fathers. San Francisco: Administrative Office ofthe Court, Judicial Council of California, 1993a. Center for Families, Children and Courts. Client Evaluations of Mediation Services: The Impact of Case Characteristics and Mediation Service Models. San Francisco: Administrative Office ofthe Court, Judicial Council of California, 1993b. Center for Families, Children and Courts. Mediated Agreements on Child Custody and Visitation. San Francisco: Administrative Office of the Court, Judicial Council of California, 1994a. Center for Families, Children and Courts. Client Feedback: Retrospective Results from the California Child Custody Project. San Francisco: Administrative Office ofthe Court, Judicial Council of California, 1994b. Center for Families, Children and Courts. Preparing Court-Based Child-Custody Mediation Services for the Future. San Francisco: Administrative Office of the Court, Judicial Council of California, 2000. Center for Families, Children and Courts. Custody Mediation and Ethnic Diversity in California. San Francisco: Administrative Office of the Court, Judicial Council of California, 2001. Center for Families, Children and Courts. Domestic Violence in Court-Based Child Custody Mediation Cases in California. San Francisco: Administrative Office of the Court, Judicial Council of California, 2002. Center for Families, Children and Courts. Difficult Cases in California CourtBased Child Custody Mediation. San Francisco: Administrative Office of the Court, Judicial Council of California, 2003a. Center for Families, Children and Courts. Court-BasedJuvenile Dependency Mediation -in California. San Francisco: Administrative Office of the Court, Judicial Council of California, 2003b. Ellis, D. Family Mediation Pilot Project. Toronto: Attorney Ceneral of Ontario, 1994.

Family Mediation Research

33

Ellis, D., and Stuckless, N. "Pre-Separation Abuse, Marital Confiict Mediation and Post-Separation Abuse." Mediation Quarterly, 1992, 9, 205-226. Ellis, D., and Stuckless, N. Mediating and Negotiating Marital Conflicts. Thousand Oaks, Calif: Sage, 1996. Emery, R. E. Renegotiating Family Relationships: Divorce, Child Custody, and Mediation. New York: Guilford Press, 1994. Emery, R. E., Matthews, S. C , and Kitzmann, K. M. "Child Custody Mediation and Litigation: Parents' Satisfaction and Functioning One Year After Settlement." journal of Consulting and Clinical Psychology, 1994, 62 (1), 124-129. Emery, R. E., Matthews, S. C , and Wyer, M. M. "Child Custody Mediation and Litigation: Further Evidence on the Differing Views of Mothers and Fathers." Journal of Consulting and Clinical Psychology, 1991, 55* (3), 410-418. Emery, R. E., and others. "Child Custody Mediation and Litigation: Custody, Contact, and Coparenting Twelve Years Aft:er Initial Dispute Resolution." Journal of Consulting and Clinical Psychology, 2001, 69, 323-332. Fabricius, W. V., and Braver, S. L. "Non-Child Support Expenditures on Children by Nonresidential Divorced Fathers: Results of a Study." Family Court Review, 2003, 41, 321-336. Folberg, J., Milne, A. L., and Salem, P. Divorce and Family Mediation: Models, Techniques, and Applications. New York: Guilford Press, 2004. Gigy, L., and Kelly, J. B. "Reasons for Divorce: Perspectives of Divorcing Men and ^omen." Journal ofDivorce, 1992, 75(1/2), 169-187. Johnson, M. P., and Ferraro, K. J. "Research on Domestic Violence in the 1990s: Ma^dn^Distmcnons." Journal of Marriage and Family, 2000, 62, 948-963. Johnston, J. R. Developing Preventative Interventions for Children of Severe Family Conflict and Violence: A Comparison of Three Treatment Models. San Francisco: Administrative Office ofthe Court, Judicial Council of California, 1994. Johnston, J. R., and Campbell, L. G. Impasses of Divorce: The Dynamics and Resolution ofFamily Conflict. New York: Free Press, 1988. Johnston, J. R., and Campbell, L. G. "A Clinical Typology of Interparental Violence in Disputed-Custody Divorces." American Journal of Orthopsychiatry, 1993, 63, 190-199. Kelly, J. "Mediated and Adversarial Divorce: Respondents' Perceptions of Their Processes and Outcomes." Mediation Quarterly, 1989, 24, 7188. Kelly, J. Mediated and Adversarial Divorce Resolution Processes: An Analysis ofPostDivorce Outcomes. Washington, D.C.: Fund of Research in Dispute Resolution, 1990. Kelly, J. B. "Parent Interaction After Divorce: Comparison of Mediated and Adversarial Divorce Processes." Behavioral Sciences and the Law, 1991 a, i? (4), 387-398. Kelly, J. B. "Is Mediation Less Expensive? Comparison of Mediated and Adversarial Divorce Costs." Mediation Quarterly 1991b, 8 (1), 15-26.

34

KELLY

Kelly, J. B. "Developing and Implementing Post-Divorce Parenting Plans: Does the Forum Make a Difference?" In C. E. Depner and J. H. Bray (eds.), NonResidential Parenting: New Vistas in Family Living. Thousand Oaks, Calif: Sage, 1993. Kelly, J. B. "A Decade of Divorce Mediation Research: Some Answers and Questions." Family and Conciliation Court Review, 1996, 34 {5), 373-385. Kelly, J. B. "Psychological and Legal Interventions for Parents and Children in Custody and Access Disputes: Current Research and Practice." Vir^nia Journal ofSocial Policy and the Law, 2002, 70(1), 129-163. Kelly, J. B., and Duryee, M. "Women's and Men's Views of Mediation in Voluntary and Mandatory Mediation Settings." Family and Conciliation Courts Review, 1992, 30 (1), 34-49. Kelly, J. B., and Gigy, L. "Client Assessment of Mediation Services (CAMS): A Scale Measuring Client Perceptions and Satisfaction." Mediation Quarterly, 1988, 19, 43-52. Newmark, L., Harrell, A., and Salem, P. "Domestic Violence and Empowerment in Custody and Visitation Cases." Family and Conciliation Courts Review, 1995,33,30-62. Pearson, J. "The Equity of Mediated Divorce Agreements." Mediation Quarterly, 1991, 7, 347-363. Pruett, M. K., and Johnston, J. R. "Therapeutic Mediation with High-Conflict Parents: Effective Models and Strategies." In J. Folberg, A. Milne, and P. Salem (eds.). Mediating Family and Divorce Disputes. New York: Guilford Press, 2004. Richardson, C. Court-Based Divorce Mediation in Four Canadian Cities: An Overview ofResearch Results. Ottawa: Department of Justice, 1988. Smyth, B., and Moloney, L. "Therapeutic Divorce Mediation: A Review." Journal ofFamily Studies, 2003, 9 (2), 161-186. Statewide Office of Family Court Services. "California Family Court Services Mediation 1991. Report 1Overview. Families, Cases, and Client Feedback." San Francisco: Administrative Office ofthe Courts, Judicial Council of California, 1992. Statistics Canada. Family Violence in Canada: A Statistical Profile, 2001. Ottawa: Minister of Industry, 2001. Thoennes, N. "An Evaluation of Child Protection Mediation in Five California Couxxs." Family and Conciliation Courts Review, 1997,35(2), 184-195. Thoennes, N. "Dependency Mediation: Help for Families and the Courts." Juvenile and Family Court Journal, Spring 2000, pp. 1422. Thoennes, N. Mediating Disputes Including Parenting Time and Responsibilities in Colorado's Tenth Judicial District: Assessing the Benefits to Courts. Denver: Center for Policy Research, Aug. 2002a. Thoennes, N. "Hamilton County Juvenile Court Permanent Custody Mediation." Denver: Center for Policy Research, Oct. 2002b.

Family Mediation Research

35

U.S. Department of Health and Human Services. Effectiveness of Access and Visitation Crant Programs. Washington, D.C.: U.S. Government Printing Office, 2002. Walker, J., McCarthy, P., andTimms, N. Mediation: The Making and Remaking of Co-Operative Relationships. Newcastle upon Tyne, UK: University of Newcastle upon Tyne, Relate Centre for Family Studies, 1994.

Joan B. Kelly, a psychologist, was director of the Northern California Mediation Center in Corte Madera, California, for twenty years. Her research, writing, and practice of over thirty-five years focused on children's adjustment to divorce, custody and access issues, and divorce mediation. She was a founding board member and former president ofthe Academy of Family Mediators. She is currently teaching seminars and writing.